Pre-Training Assessment: Free Template and Example Items

Train teams faster and more effectively with AI

Start Free

Picture fourteen people starting on an electrical wholesaler's trade counters next month. Six already know electrical products from a business the company has recently bought, five know counter work from hardware retail, and three know the stock from the warehouse but have never sold anything. They all face the same 150-minute course. Before it opens, a pre-training form asks each of them "How confident are you checking stock?" on a scale of 1 to 5, and the course goes ahead the same way for everyone whatever they answer.

I'm an instructional designer at Coursebox, which makes an AI quiz generator that drafts questions from your own training documents, so I'll say where the product helped and where it needed a correction. In this guide I explain what a pre-training assessment is, why a confidence survey cannot do its job, and how to write items that test the task, mark them and decide who skips what. The guide works through that trade counter as a worked example with 8 items and a mark scheme, and comes with a free template, real cases from New Zealand, the US and Ghana, and a product test.

What is a pre-training assessment?

A pre-training assessment is a short test, taken before training starts, that checks what each learner can already do in the tasks the training covers, so you can decide who skips, shortens or takes each part.

You will also see it called a pre-assessment, a pre-training test or a pre-course assessment. In education it is usually filed under diagnostic assessment, one of the types our guide to formative and summative assessment sets out.

A pre-training assessment does two jobs:

• It sets each person's path. Someone who shows they can already do a part of the job reads a summary instead of the full module. Someone who cannot takes all of it.
• It records a starting point. If the same items come back after training, the change between the two is a before and after measure for each person and each item.

If a result will not change anyone's path or feed a later comparison, it is a quiz with no purpose. Decide what each result will change before you write a single item.

Why test who is in the room first

A course designed for beginners does not stay neutral for people who already know the material. Sweller, van Merriënboer and Paas put it plainly in their review of cognitive load research: "Instructional procedures designed for novices dealing with multiple, interacting elements can be counterproductive as expertise increases." Most of that evidence comes from schools and universities, but the logic carries to a mixed group of new staff. Sending the experienced people through the novice version costs their time and gives them little in return.

A government evaluation in New Zealand shows what happens when a pre-test finds this and nobody acts on it.

Case study: New Zealand employers found staff already at the target level

From July 2006 to June 2009, New Zealand's Department of Labour funded workplace literacy courses through 15 partnerships with employers in industries including roading, meat and seafood processing, manufacturing, hospitality and healthcare. Employers, not a test, chose who went on the courses.

• What they did: Every participant was interviewed and given a formal reading and writing test before the course. Only those below the target level were re-tested afterwards, on a different version of the test.
• Evidence and result: The Upskilling Partnership Programme report states that 52 of 334 learners in that pre-course analysis were already at IALS/ALL level 3 or above, the target level. That is 15.6%, about one in six. The evaluators questioned whether those places could have gone to other employees and reported strong criticism from a disproportionate number in this group. Among 280 re-tested learners, reading scores rose by 10.1 points, but scores were not related to attendance hours.
• Limitation: There was no control group, because one company allocated workers by need instead of at random. The authors themselves doubt the size of the test gain, workplace changes were rated by supervisors, and the government ran the evaluation of its own programme.
• Lesson: The pre-test identified learners already at the target level, but the information did not change attendance. Decide in advance who will review the evidence and what they can change. A target reading level alone does not establish that every workplace learning need has been met.

Pre-training assessment vs survey, needs analysis and knowledge check

Four instruments get confused before training starts, and each answers a different question:

A survey about expectations, access needs and logistics can sit beside the assessment, and our list of training survey questions covers that instrument. A training needs analysis questionnaire works at the organisation and role level, to decide what training to build in the first place. For a team's skill gaps and their causes, use a skills gap analysis template. Checks during the course are covered in our guide to e-learning knowledge checks.

One more boundary for Australian readers: recognition of prior learning is a formal assessment process for nationally recognised training, run by a registered training organisation, as ASQA explains. An internal pre-training assessment borrows the idea of credit for what someone can already show, without that regulated process.

Why a confidence survey is not a pre-training assessment

Many pre-training "assessments" are confidence surveys: rate your knowledge of returns, rate your confidence with pricing, 1 to 5. The problem is that self-ratings say more about how people feel than about what they can do.

The clearest evidence on this comes from a meta-analysis of self-assessed knowledge in education and workplace training by Traci Sitzmann, Katherine Ely, Kenneth Brown and Kristina Bauer. They pooled 166 studies covering 41,237 learners and compared people's ratings of their own knowledge with tests of what they had learned.

A correlation of .34 is a moderate relationship, not "34% accurate". The authors' own summary is that "The relationship between self-assessment and cognitive learning was moderate", while self-ratings had their "strongest correlations" with motivation and satisfaction. Self-ratings track how people feel about training more closely than what they know.

The figure moves with how the question is asked and how the course is delivered (pp. 177 to 178). Against the overall .34:

• "How knowledgeable are you about this?" Ratings of current knowledge correlated .44 with measured learning, the best case of any question type.
• "How much did you learn?" Ratings of how much someone had gained correlated .00, no relationship at all.
• Delivery: In web-based courses the correlation fell to .19.

Two limits apply. Three in four studies were of university students, and most of the ratings were collected at the end of a course, not before it. The logic carries to a pre-course form, but none of these numbers was measured on one.

The authors spell out the practical risk themselves. Basing course decisions on self-ratings "may lead organizations to overlook areas of knowledge deficiency", and they suggest "short knowledge tests made available on-line" as a way to find the topics employees need training on. They also found that "nearly a third of evaluation studies interpreted self-assessed knowledge data as evidence of cognitive learning".

Here is the same question asked both ways, for new staff on the trade counter of an electrical wholesaler, who have to find stock, quote prices and handle returns by the business's own rules:

The left column gives you a number about how someone feels. The right column gives you an answer you can mark against a rule.

A worked example: the trade counter

The rest of this guide follows one assessment from first draft to results. The example is hypothetical; the business, the people and the results are illustrative.

An electrical wholesaler in South East Queensland has 5 branches: Eagle Farm (head office and training room), Caboolture, Kunda Park, Molendinar and Beenleigh. It sells cable, switchgear, lighting and data products over the trade counter, mostly to electricians on trade accounts, with some walk-in customers. It has recently bought an independent wholesaler in Beenleigh, which is now its fifth branch.

Fourteen people start on the trade counters over the next month:

• 6 from the Beenleigh business: they know electrical products and trade customers, but not this company's system or its account, order and return rules.
• 5 from hardware and building-supplies retail: they know how to run a counter, but not electrical products or trade accounts.
• 3 from the Eagle Farm warehouse: they know the stock, the part numbers and the stock screens, but have never priced or sold.

Mereana, the training coordinator at Eagle Farm, writes and marks the assessment. The course is "Trade counter essentials", 150 minutes in four modules:

The course teaches the business's own 8 counter rules, not wiring rules, electrical safety or consumer law:

1. Out of stock here: before saying no, check the other four branches' stock and the incoming purchase orders. An inter-branch transfer ordered by 2 pm arrives the next business day.
2. Superseded part numbers: the item record shows "Replaced by". Offer the replacement and check its specification with the customer before selling it.
3. Account pricing: quote account customers the price level on their account, never the shelf price, and save every quote against the account.
4. Quote validity: quotes are valid for 30 days, except copper cable, which is valid for 7 days because supplier cable prices move with the copper price.
5. Special orders: customers without a trade account pay a 50% deposit before a special order is placed, and every customer is told special orders cannot be returned unless faulty.
6. Substitutions: if a supplier offers a different brand on a special order, accept it only with the customer's approval, recorded on the order.
7. Returns: unused items in their original packaging, within 30 days, with the invoice number, are credited to the account or refunded to the original payment method.
8. Cut cable: cut lengths cannot be returned for a change of mind. If the customer says the cable is faulty, start a warranty claim form; never refuse a fault claim at the counter.

This example shows a pre-training assessment well for three reasons. The three groups know different parts of the job, so one course for everyone wastes time in three different places. The gaps differ by module as well as by person. And every rule is written down, so every answer can be marked.

How to write pre-training assessment items

A good item puts the learner in a real moment from the job and asks what they would do. The clearest rules for this come from medical licensing exams. The NBME Item-Writing Guide says "Each item should assess application of knowledge, not recall of an isolated fact" (p. 33), and I apply its rules to workplace items like this:

• Ask for an action or a decision. "What do you do?" tests the job. "What is the rule for copper cable?" tests memory.
• Use a real situation from the job. A customer, a request and a detail that changes the answer, such as "plans to order in about three weeks".
• Make it markable. One best answer, or a short written answer you can mark against a written rule.
• Use the cover-the-options test for a focused multiple-choice item: a knowledgeable learner should be able to anticipate a suitable answer before seeing the alternatives. This checks whether the question is focused, not whether recognition is inherently invalid. Include recall questions when accurate recall is itself a job requirement.
• Leave out "usually" and "often". NBME's reason is that such terms "are interpreted differently by different people" (p. 25).
• Check that the course answers it. Every item needs a page in the course that teaches the answer, or the assessment can route someone to a module that never helps.

The 8 items and their mark scheme

Mereana writes two items per module, each marked 2, 1 or 0. A module score is out of 4, and the whole assessment is out of 16.

The 1-mark column records an incomplete answer so the trainer can target the missing part. It does not excuse the error: a learner who gives the wrong cable-quote period still studies that rule and demonstrates it correctly before relying on it at the counter.

The template has these 8 items, the mark scheme, the bands and a results sheet filled in, with blank pages for your own course.

If you draft items with an AI tool, supply the approved rules, required task and representative situations. Inspect every answer and explanation. Use application questions where decisions are required, and direct recall where people need a fact from memory. An item's format alone does not establish what it measures.

How long it should take, and when to send it

• Keep it short. Eight items take about 15 minutes. In a randomised trial with hospital residents, the researchers built their questions into the modules on purpose, because a separate pretest in another study had added 30 minutes to the session and could have cancelled out the time saved.
• Send it close to the start. Each of the 14 takes it on their first morning, before the course opens, and Mereana marks it by lunchtime.
• Say what it is for. Use a clear introduction: "This assessment helps Mereana plan the parts of the course you need. It is not a course grade or a qualification decision. She will explain your path and check any uncertain answers with you."
• Keep it ungraded. The result sets a path. It is not a pass or a fail, and it should not count towards anything.
• Show the path, not the answer key. The course pages supply the answers.

Attempting the items before the course has a side benefit. Attempting a question before studying tends to help people learn the answer to that question, as long as they get the correct answer afterwards, according to Pan and Carpenter's review of pretesting research. The same review calls the evidence for material that was not asked about "highly inconsistent", and most of its participants were undergraduates. So the priming covers only what you ask, which is one more reason to write items on the rules that cause the most trouble at the counter.

How to read pre-training assessment results

Decide what each score will change before the assessment goes out. Otherwise the results arrive, and the easiest thing to do is send everyone through everything anyway.

Three bands for each module

Mereana bands each module on its own score, never on the total out of 16. A total of 10 can hide a 0 on Price it, and a new starter with a 0 on Price it will quote the wrong price no matter how well they know returns.

The bands below are illustrative planning rules for these sampled counter tasks, not validated cutoffs or proof of competence. Pilot the items and mark scheme, confirm coverage with a subject expert, and review uncertain answers before shortening content. Treat blank or unobserved responses as missing evidence, not a silent zero or a pass; arrange an accessible re-attempt. Required induction and workplace sign-off remain.

Two readings of the whole group

Before sending anyone their path, look at the group as a whole:

1. Nearly everyone lost marks on the same item. If 10 or more of the 14 lose marks on one item, treat it as a problem with the item or with what people could have known, not as 10 weak learners. Reread the wording, then ask whether the rule was ever available to them. If the rule was never available, teach it to everyone before the course starts; if the wording was the problem, fix it before the next group.
2. Scores split by background. Look for a possible shared gap, then inspect each learner's item responses and access needs before assigning a path. A group pattern can guide support; background is not a substitute for individual evidence.

Here are illustrative results for the 14:

Both readings show up in these results. The first is a single item:

• The item: 13 of the 14 lost marks on item 4, the copper cable quote.
• The reading: The 7-day rule is this business's own, so people from outside are unlikely to know it, and the warehouse staff have never quoted. That is a rule to teach everyone, not evidence about 13 people.
• The change: Mereana adds the 7-day rule to the first-morning briefing for all 14. The wording of the item stays as it is.

The second is a split by background:

• The group: All six from Beenleigh took the short or full path on Find it, but four of them skipped Order it and two skipped Return it.
• The reading: They know special orders and returns from their old business. What they lack is this company's system.
• The change: Mereana books a 20-minute system walkthrough with an experienced counter hand from Eagle Farm on day 1 and keeps them on Find it.

Two individual paths

• Shreya, from hardware retail: Find it 1 (full, 40 minutes), Price it 1 (full, 40 minutes), Order it 1 (full, 35 minutes), Return it 3 (short, 15 minutes: lost a mark on item 8). Planned course time: 130 minutes instead of 150.
• Tane, from the Beenleigh business: Find it 1 (full, 40 minutes), Price it 2 (short, 20 minutes), Order it 4 (skip, 5 minutes), Return it 4 (skip, 5 minutes). Planned course time: 70 minutes instead of 150.

These are planned minutes, not a result. Nothing in the example shows that either of them works better at the counter because of it.

Branching learners onto their paths

Letting people skip what they can already do saves time, and the best test of whether it costs learning comes from a hospital.

Case study: Mayo Clinic residents skipped content they already knew

Internal medicine residents at Mayo Clinic in Rochester, Minnesota, are employed doctors in training, and a set of web modules on diabetes, high cholesterol, asthma and depression was a required part of their clinic curriculum. The authors wrote that "Increased clinical demands and decreased available time accentuate the need for efficient learning".

• What they did: Each module came in two versions. In the adaptive version, a resident who correctly answered a case question could "skip the corresponding content"; a wrong answer meant working through it. The questions were built into each module, not given as a separate pretest. Each resident did two modules in each version.
• Evidence and result: In Cook and colleagues' 2008 randomised crossover trial, 124 residents were randomised and 122 completed at least one module. Knowledge scores were about the same (76.2 adaptive, 77.2 standard), and the adaptive version took 29.3 minutes per module against 35.6, "an 18% decrease in time". Of 96 residents who answered the preference survey, 72 (75%) preferred the adaptive format.
• Limitation: One programme at one institution, about 6 minutes saved per module, and knowledge measured straight after each module, not in practice with patients. The study had no pre-test and no group without training, which the authors note leaves open whether any learning happened at all.
• Lesson: The adaptive format saved about 18% of module time, with no statistically reliable difference detected on the immediate knowledge test. That is not an equivalence or non-inferiority result, and it does not establish unchanged patient-care performance. Keep diagnostic questions short enough that their time cost does not consume the saving.

The clearest published evidence on letting staff test out comes from hospitals, and none of it shows job performance changing because of a pre-test. Ohio State University's medical centre let clinicians who already knew its records system "test out of Basic Content" in their training, but the in-house team's report measured its 45% faster training across the whole move from classroom to self-paced eLearning, so that figure cannot be put down to the test-out alone.

Setting up the paths in practice

Branching sounds like software, but at this scale it is a decision a person makes and a course built to allow it. Here is how the trade counter course handles it:

• Who decides: Mereana reads each person's module scores against the bands and gives them their path in writing, such as "Find it: full. Price it: pages for item 4. Order it and Return it: rule summaries".
• How the course allows it: Each module opens with its one-page rule summary, and each rule has its own page, so a short path is a list of pages and a skip is one page.
• How learners move through it: In Coursebox, you can leave a course's sections open so learners jump to the pages on their path from the left-hand menu. You can also split modules into separate courses and chain them with prerequisites or a learning pathway. Either way, a person assigns the path; the course does not read the score and send anyone anywhere.

Two questions sent to Coursebox support are about exactly this step. One training company was considering a short pre-assessment to move experienced practitioners past beginner content. A training college wanted to compare pre-test and post-test results to show learning gain. Both needed a person to do something with the scores, so plan who reads them, who assigns paths and where the comparison will be done before you send the assessment.

If you are building the modules themselves, our page on AI for employee training shows how Coursebox drafts the outline, lessons and quiz questions from your own documents.

Using the same items after training

Using identical items makes the content of a before-and-after comparison easy to align. In this example, learners answer the eight questions again at the end of week 2 and Mereana uses the same marking scheme. This checks those answers, not overall competence. Equivalent new situations can also help check application, but their difficulty and coverage need to be comparable. The knowledge comparison belongs at Level 2 of the Kirkpatrick model.

Read the change item by item as well as in total. A study of health workers in Ghana shows why. Paintsil and colleagues gave the same 20 questions before and after a one-day workshop on sickle cell disease, for health workers from 46 facilities. Among the 210 people with both tests, the average score rose from 8.2 to 12.9 out of 20. Across the study, the share failing the test fell from 65% to 21%. One topic did not move: on mental health and psychosocial support, 4.58% answered correctly before and 4.07% after. The total went up, and the item showed which part needed redesigning. The study had no control group and tested people straight after the workshop, so it shows knowledge on the day, not practice.

Two cautions apply to any comparison on identical items:

• Seeing and attempting the same items can contribute to a later rise, especially once the correct information has been taught. The amount cannot be separated here from instruction, other experience and measurement effects. Do not attribute the whole change to the course.
• A written answer is not the job. The assessment shows someone can say what to do at a desk. In the trade counter example, each person also does 3 observed counter shifts with their branch manager, signed off against the 8 rules. That is where they show they can do it with a customer waiting.

Building the assessment with an AI course tool

On 26 October 2026 I briefed the Coursebox course agent to build "Trade counter essentials" with the pre-training assessment at the front. Coursebox is our product, so weigh my account with that in mind. The brief gave the 8 counter rules word for word and the 8 situation questions in order, asked for a "Before you start" page explaining the three bands for each module, and said that Mereana marks the assessment and gives each person their path. It also asked for the same 8 questions at the end of week 2.

I expected it to miss the items themselves: to turn the situation questions into recall multiple choice, or to add a "how confident are you" question at the start. That did not happen. The 8 questions came back word for word, in order, as short-answer prompts, starting with "An electrician asks for 3 boxes of the 10 W tri-colour downlight. Your branch shows 0 in stock. What do you check before you answer, and what can you offer?" No confidence or self-rating question was added. The page said what the assessment was for and who marks it: "This decides which parts of the course you can skip. It does not count for anything. Mereana marks it manually and gives each person their path through Find it, Price it, Order it and Return it. No result, correct answer or answer feedback is shown after submission." It promised no automatic routing, and the block itself said "This diagnostic is ungraded and is marked manually by Mereana."

The problem was in the block's marking setting. Under "How this is marked" it read "AI marks instantly", and in the block's Marking panel, "Who marks it?" was set to "AI marks instantly: Learners get a grade and feedback as soon as they submit." So a new starter would have received an instant AI verdict on a diagnostic that the page said shows no result and counts for nothing, before Mereana had seen it.

In the Marking panel I changed "Who marks it?" to "A facilitator marks: Submissions wait in the grading queue for a person." After I reloaded the page, the block read "A facilitator marks", and submissions now wait for a person.

After that change, the block still said learners pass when a submission is marked Satisfactory. That pass wording was not changed. All eight answers shared one submission box, and no model answer was stored in the block; Mereana would apply the per-item 2, 1 and 0 scheme herself. The page's claim that the diagnostic counted for nothing also needed clearer wording, because it determined a learning path. The run did not resolve those points or test an actual submission moving through marking.

I inspected the outline, the pre-assessment page and its marking setting after a reload. I did not enrol learners, mark answers or assign paths. Coursebox holds the course and each person's submitted answers; Mereana marks them and assigns each path. Nothing here shows that the assessment sorts people correctly or that anyone serves customers better because of it.

Putting a pre-training assessment to work

A pre-training assessment is worth the time it takes only when it tests the job, not people's confidence, and when someone acts on every result. Self-ratings track how people feel more closely than what they know, so ask each learner to handle a real situation, mark the answer against your own rules, and set each module's path from that module's score. Then read the group before the individuals, because an item everyone misses is usually telling you about the item or about a rule nobody was ever taught.

Your next step: take the course you run most often and write one situation item for each module, with what earns 2, 1 and 0 marks. Before you send it, write down what each band will change for that module. If no band changes anything, leave that module out of the assessment.